The default setup is the problem
Most retargeting is configured once: everyone who visited the website in the last thirty days, one creative, no cap, no exclusions. It runs for a year.
That configuration produces the experience everybody complains about — the same product following someone across the internet for weeks, including the week after they bought it.
Segment by what they did
A person who read a blog post is not a person who abandoned a cart. Build separate audiences:
- Content readers. Low intent. Show brand or educational content, not a hard offer.
- Service or product page viewers. Moderate intent. Show proof, comparison, and reasons to trust.
- Enquiry or cart starters who did not finish. High intent. Address the specific friction — cost, delivery, trust — and make completion easy.
- Past customers. Different objective entirely: repeat purchase, cross-sell, referral.
The message that works for each is completely different, and a single campaign speaking to all of them speaks properly to none.
Match the window to the buying cycle
- Fast consumer purchases: short windows, days.
- Considered services: several weeks.
- High-value purchases such as property, B2B systems, or education: months, with content that respects the length of the decision.
A thirty-day default is arbitrary. Use your own data — the average time between first visit and purchase — to set it.
Cap frequency deliberately
Effect flattens and irritation rises quickly. Set caps at a level you would tolerate as a consumer, then check reach and frequency reports monthly rather than setting and forgetting.
Signals you are over-serving: frequency climbing while click-through falls, negative comments on ads, and rising cost per result in a stable market.
Exclusions are not optional
Always exclude:
- People who converted, immediately upon conversion.
- Existing customers, from acquisition campaigns.
- Job applicants and current employees, where identifiable.
- Anyone who asked to stop.
Paying to advertise a product to the person who bought it yesterday is both a waste and a small insult, and it is startlingly common.
Sequence the message
Rather than repeating one creative, move the story forward:
- First few days: the reason to reconsider — a benefit, a specific proof point.
- Next phase: address the most common objection directly, whether that is price, trust, or timing.
- Later: a reason to act now — availability, season, an offer if appropriate.
- Final phase: reduce frequency substantially and switch to brand-level presence.
Creative that works in retargeting
- Real proof: customer footage, delivered work, reviews.
- Specific answers to specific hesitations.
- Short, direct, without a long build-up — the audience already knows who you are.
- Fresh executions every few weeks, because this audience is small and sees everything.
Privacy and permission
Tracking capability continues to narrow through browser and platform changes, and regulation continues to tighten. Practical response:
- Rely more on first-party data — your own list, your own customer records.
- Implement consent properly rather than as an obstacle course.
- Use server-side measurement where appropriate.
- Assume audience sizes will shrink and value each visitor more.
Measuring incrementality honestly
Retargeting reports look excellent because they take credit for people who were already going to convert.
Test it: hold out a portion of the audience for a period and compare conversion rates. If the difference is small, you are paying to reach people who would have returned anyway.
This test is uncomfortable and worth running annually. It frequently reallocates meaningful budget toward acquisition where the incremental effect is larger.
The principle
Retargeting works when it feels like a helpful reminder from a business the person was already considering. It fails when it feels like surveillance.
Segmentation, caps, exclusions, and message progression are the difference — and all four are configuration decisions, not budget decisions.
